Skip to content
Skills Library
AI & agents · Automation

gauntlet-loop

/gauntlet-loop

Turns any goal into one short, paste-ready gauntlet-loop prompt for GitHub Copilot CLI, then can run it with the task tool and /fleet for disjoint pieces. No /loop, no ultracode — those are Claude Code. Triggers on "/gauntlet-loop", "gauntlet loop", "gauntlet this".

Licence
CC-BY-4.0
Last repository change
04.09.2026

Assessment and security

95/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content30 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
3
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 04.09.2026, without references to retired models or APIs.

Licence

CC-BY-4.0, otwarta.

What it does and when to use it

Turns any goal into one short, paste-ready gauntlet-loop prompt for GitHub Copilot CLI, then can run it with the task tool and /fleet for disjoint pieces. No /loop, no ultracode — those are Claude Code. Triggers on "/gauntlet-loop", "gauntlet loop", "gauntlet this".

Triggers from the description

gauntlet loopgauntlet this

Arguments

/gauntlet-loop <goal>
Original author description

Turns any goal into one short, paste-ready gauntlet-loop prompt for GitHub Copilot CLI, then can run it with the task tool and /fleet for disjoint pieces. No /loop, no ultracode — those are Claude Code. Triggers on "/gauntlet-loop", "gauntlet loop", "gauntlet this".

How it works

  1. Prompt template
  2. Run it here
  3. Example (visual)
Original SKILL.md headings · 51 lines, 564 words
SKILL.md preview
---
name: gauntlet-loop
description: >
  Turns any goal into one short, paste-ready gauntlet-loop prompt for GitHub
  Copilot CLI, then can run it with the task tool and /fleet for disjoint
  pieces. No /loop, no ultracode — those are Claude Code.
  Triggers on "/gauntlet-loop", "gauntlet loop", "gauntlet this".
compatibility: copilot-cli
license: CC-BY-4.0
argument-hint: "<goal>"
---

# Gauntlet Loop (Copilot CLI)

You are on **GitHub Copilot CLI**. Spawn with `task`. Independent file-disjoint pieces may use `/fleet`. File tools are `view`, `create`, `edit`, `apply_patch`, `bash`.

**No `/loop`. No `ultracode`.** Those are Claude Code.

Read `<this-skill-dir>/references/loop.md` for bar tests, flow, and voice. Read `references/run-copilot.md` for spawn details.

Apply the evidence and completion contract in `references/loop.md`: select references before implementation, isolate blind A/B critic packets, inspect the shipped experience, and require passing checks plus the mode-specific completion condition. Preserve these requirements in the generated prompt.

## Prompt template

```
Build [GOAL].

The bar is [BAR]. Get the real thing first — fetch it, screenshot it, read it, or run it — and compare against that artifact, not against a memory or a description of it.

You are the lead. Break this into the smallest pieces that can be improved and judged on their own. For each piece, spawn a builder with the task tool and a separate critic with the task tool and fresh context. Independent pieces may run in parallel or via /fleet. Same-file pieces stay serial. A critic is a new task every time.

The critic inspects the actual output, puts it next to the bar with the labels stripped, says which one is better, and names the single biggest remaining gap. Then you send only that gap back to the builder. The critic should be harsh. Praise is not useful. A score is not useful. If ours does not win, keep going.

Keep looping until the critic picks ours blind — or, for a bar in a different medium than the product, until a challenger version stops beating the champion and no movable gap remains. Park any piece that loses twice on the same gap and report it. Do not stop before that. Do not name a round count.

Keep a live progress file updating as the work evolves so I can watch it.

Spawn builders and critics with the task tool. There is no /loop and no ultracode in this session.
```

## Run it here

Follow `references/run-copilot.md`.

## Example (visual)

```
Build a landing page for a running brand. Athletic, peak performance, green and dark, energetic, aimed at a young healthy audience. It needs to be interactive and visually unmistakable.

The bar is Nike's current running campaign page. Open it in the browser, screenshot desktop and mobile, and compare against those images directly, not against a description of them.

You are the lead. Break this into the smallest pieces that can be improved and judged on their own — hero, motion, type, colour, imagery, interaction, mobile. For each piece, spawn a builder with the task tool and a separate critic with the task tool. Independent pieces may use /fleet. Never let a critic edit the product.

The critic opens the real Nike page and our page, puts the screenshots next to each other blind with the labels stripped, says which is better, and names the single biggest remaining gap. Then you send that gap back to the builder. The critic should be harsh. If ours does not win, it keeps going.

Keep looping until the critic picks ours blind. Do not stop before that.

Keep a live progress file updating as the work evolves so I can watch it.

Spawn builders and critics with the task tool. There is no /loop and no ultracode in this session.
```
Files in this package (3)
  • SKILL.md 4 KB
  • references 1 KB
  • workflows 1 KB

Versions

RepositoryScoreLicence
This page · aaronwjordan/gauntlet-loop95CC-BY-4.0
emaraschio/cursor-commands100MIT
mintuz/claude-plugins100MIT
robonuggets/gauntlet-loop100CC-BY-4.0
aaronwjordan/gauntlet-loop95CC-BY-4.0
aaronwjordan/gauntlet-loop95CC-BY-4.0
aaronwjordan/gauntlet-loop95CC-BY-4.0
rubenmarcus/agent-skills90MIT
aaronwjordan/gauntlet-loop90CC-BY-4.0

Identical content appears in 1 other GitHub folders. We show one copy with the highest score.

Database snapshot: 13.11.2026 · package version 7352164f6d9e (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100setup
    Set up a Gauntlet Loop for a project: define the goal, the reference bar the work must beat, judging criteria beyond looks, how each piece is captured and judged (the harness), and the stop rules, then write and validate a gauntlet.json spec. Use for 'gauntlet this', 'set up a…
  • 100run
    Run a Gauntlet Loop from a gauntlet.json spec: split the work into pieces, build each with a fresh hotline callee, capture the real output, judge it blind against the reference with an independent critic callee, commit each round, and loop until it wins, a stop rule fires, or…
  • 90gauntlet-performance
    Use when a human asks how gauntlet runs perform across the recorded telemetry - explicit invocation only (/skill:gauntlet-performance [--dir <path>]... [--since <version>]).
  • 90gauntlet-proof-audit
    Use to audit the gauntlet records, "are the ROUNDS.md records honest", "verify the proof corpus", "check each run record against its evidence", "audit the gauntlet corpus", "is the gauntlet proof trustworthy", or a scheduled run. NOT for re-running probes or rounds…
  • 100gauntlet-loop-brain
    Design and govern evidence-bound Gauntlet Loops: one initiating prompt that launches coupling-aware builders, fresh critics, protected gates, integration checks, and explicit stop controls. Use for Gauntlet fit decisions, job contracts, reference selection, builder-critic loops,…
  • 100gauntlet-loop-max
    Two-gate quality loop for work that tests alone can't prove is good - UI, animation, dashboards, landing pages, copy, scripts, titles, thumbnails, docs, decks. Gate 1 proves it works (spec + quality review by an independent reviewer, fix rounds, severity-ranked findings). Gate 2…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.